Machine Learning Study Goals – October 2022: Time Series

Machine Learning Study Goals October with Jack O Lantern Background and Sarah's Selfie in the corner

Welcome to another installment of: I share my goals for improving at Machine Learning and we can all guess how many of those I will reach 😉 I will post an update at the end of the month as always.

My theme for this month is time series. Last month I had some time series related goals that I didn’t reach and this time I’m putting it into focus and starting from zero to give it a proper shot.

Table of Contents

Motivation: Personal curiosity and improving my work skill set

Time series and dealing with sequential data in general is a weak spot for me and I want to be able to at least know the basics in case I get confronted with such a problem at work (I work as a Data Scientist in IT consultancy) – and because of curiosity too. I also hope that having some knowledge and experience with time series will help me out later down the line when I tackle transformers and NLP again.

Time series data is very common for retail industries, which are often clients of ours. It could be predicting upcoming sales to order approriate amounts, but it can also be used to detect faulty data in the data pipeline by detecting outliers before working further with the data.

But enough about my motivation, let’s get into the concrete goals:

Goal 1: Read “Machine Learning for Time-Series with Python”

I’m talking about this book here by Ben Auffarth:

Mind you, I don’t want to read the whole book. I’m aiming for chapter 2 and 3, since chapter 1 is simply an introduction of which industries use time-series data.

These chapters deal with the Python basics of which libraries to use and pre-processing of the data.

It was quite hard to find recommended books on this topic, so I hope this resource will be worth the time reading it, but I do want to take my time and read a book instead of using exclusively online resources – because I always distract myself online…

Goal 2: Time Series with Python track on DataCamp

I bought a one-year plan for DataCamp at the beginning of the year and haven’t used it in a while, so I want to check out their content on time series for days when I don’t feel like reading a book.

This is advertised as a 20-hour-course, so this will take up a good chunk of my study time for the month.

Bonus Goal: Do 1 time series analysis & post it

Since Goal 2 will be quite a big one in terms of time investment, I will declare this as a bonus goal.

But I think it would be interesting to take a small dataset and apply what I learned this month to it and share it with you here. I can then later improve on it when I learn more and see how I improved.

Update: Results of my studies coming at the end of the month

Sooooo, I have bad news. Or maybe realistic news. Because this month I did not reach any of my goals from this post. After my fairly successful previous months, this had to happen at some point. I think there were two factors to this:

  1. it was a rough month at work. We’re going into a new project and the start was everything but smooth, and while things are under control and on a good trajectory now, the situation was just stressing me out as someone who likes to have a plan and feel in control.
  2. My goals weren’t exciting. Reading books and doing courses in this very structured way did not inspire me. It wasn’t fun. Also I can’t really recommend either of the two resources I mentioned above sadly…

So for next month I’m prioritizing project-based goals and exploration around a topic instead of strictly following a book. Lesson learned.

I did publish a time series blog post though: Time Series Analysis and Forecasting – An Introduction

Previous month’s study goals & results

July 2022: Decision Trees

August 2022: Random Forests

September 2022: Transformers

Leave a Reply

Consent Management Platform by Real Cookie Banner